Indirect Effects of a Fungal Entomopathogen, Lecanicillium lecanii (Hypocreales: Clavicipitaceae), on a Coffee Agroecosystem Community Author(s): A. J. MacDonald, D. Jackson, and K. Zemenick Source: Environmental Entomology, 42(4):658-667. 2013. Published By: Entomological Society of America URL: http://www.bioone.org/doi/full/10.1603/EN12287

BioOne (www.bioone.org) is a nonprofit, online aggregation of core research in the biological, ecological, and environmental sciences. BioOne provides a sustainable online platform for over 170 journals and books published by nonprofit societies, associations, museums, institutions, and presses. Your use of this PDF, the BioOne Web site, and all posted and associated content indicates your acceptance of BioOne’s Terms of Use, available at www.bioone.org/page/terms_of_use. Usage of BioOne content is strictly limited to personal, educational, and non-commercial use. Commercial inquiries or rights and permissions requests should be directed to the individual publisher as copyright holder.

BioOne sees sustainable scholarly publishing as an inherently collaborative enterprise connecting authors, nonprofit publishers, academic institutions, research libraries, and research funders in the common goal of maximizing access to critical research. COMMUNITY &ECOSYSTEM ECOLOGY Indirect Effects of a Fungal Entomopathogen, Lecanicillium lecanii (Hypocreales: Clavicipitaceae), on a Coffee Agroecosystem Ant Community

1,2 3 3,4 A. J. MACDONALD, D. JACKSON, AND K. ZEMENICK

Environ. Entomol. 42(4): 658Ð667 (2013); DOI: http://dx.doi.org/10.1603/EN12287 ABSTRACT Fungal entomopathogens are widely distributed across natural and managed systems, with numerous host species and likely a wide range of community impacts. While the potential for fungal pathogens to provide biological control has been explored in some detail, less is known about their community interactions. Here we investigate the effects of fungal epizootics of the entomopatho- gen Lecanicillium lecanii (Zimmerman) on a keystone mutualism between Azteca instabilis (F. Smith), a dominant arboreal ant, and the green coffee scale (Coccus viridis Green), as well as broader impacts on a coffee agroecosystem ant community. We hypothesized that seasonal epizootics cause shifts in the foraging ranges of A. instabilis as the adapt to the loss of the resource. We further hypothesized that the magnitude of these shifts depends on the availability of alternate resources located in neighboring shade trees. To test these hypotheses, we induced an epizootic in experimental sites, which were compared with control sites. Surveys of ant activity were undertaken pre- and post- epizootic. We found a decrease in foraging activity of A. instabilis and increase in activity of other ant species in the experimental sites post-epizootic. The decrease in abundance of A. instabilis foragers was greater on plants in which an epizootic was induced than in other plants. This relationship was modiÞed by shade tree density where higher shade tree density was associated with larger decreases in A. intabilis foraging activity in coffee plants. These results demonstrate the potential for fungal entomopathogens to inßuence the structure and diversity of ecological communities.

KEY WORDS Lecanicillium lecanii, fungal epizootic, Azteca instabilis, self-organization, disturbance

Fungal entomopathogens have been widely studied as cesses that structure communities is an important av- biological control agents in numerous systems and enue of research. against numerous different pests (Shah and Pell 2003, Community structure and patterns of diversity have Scholte et al. 2005, Pell et al. 2010). While the direct been demonstrated to result from numerous processes effects of the use of fungal biocontrol on pest species and underlying biotic and abiotic conditions (Paine have been of primary interest to date, the important 1966, 1969; Connell 1978; Hastings 1988; Caley and St. roles these organisms play as interacting members of John 1996; Chesson 2000; Hansen 2000; Bascompte and ecological communities in complex food webs are Jordano 2007; Perfecto and Vandermeer 2008). Of poorly understood and little studied (Meyling and particular importance in our study system, spatial pat- Hajek 2010). Population regulation and host-pathogen tern has emerged as an important variable in the study interactions are implicit in the treatment of fungal of the distribution of biodiversity at numerous scales entomopathogens simply as agents of biological con- (Ritchie 2010). The spatial distribution of a population trol, yet indirect effects with important implications or community throughout a landscape has often been for the structure and diversity of communities are assumed to result from underlying habitat structure omitted. Therefore, addressing the role of ento- mopathogenic fungi in complex interactions and pro- (Forman and Gordon 1987, Turner et al. 1990). How- ever, there may be other factors contributing to ob- 1 School of Natural Resources and Environment, University of served spatial patterns, particularly when considering Michigan, 3531 Dana Bldg., 440 Church St., Ann Arbor, MI 48109. relatively uniform landscapes (Skarpe 1991, Rohani et 2 Corresponding author: Department of Ecology, Evolution, and al. 1997, Bascompte and Sole´ 1998, Alados et al. 2007, Marine Biology, UC Santa Barbara, 2107 Noble Hall, University of Scanlon et al. 2007, Sole 2007). For example, observed California, Santa Barbara, CA 93106 (e-mail: andrew.macdonald@ ´ lifesci.ucsb.edu). patterns and distributions in relatively uniform eco- 3 Department of Ecology and Evolutionary Biology, University of systems, such as managed agroecosystems, may be the Michigan, 2077 Kraus Natural Science, 830 North University, Ann result of endogenous self-organization arising from Arbor, MI 48109. 4 Ecology Graduate Group, UC Davis, Department of Entomology, species interactions rather than underlying habitat 320 Briggs Hall, One Shields Av., Davis, CA 95616. variables (Pascual et al. 2002, Sole´ and Bascompte

0046-225X/13/0658Ð0667$04.00/0 ᭧ 2013 Entomological Society of America August 2013 MACDONALD ET AL.: ENTOMOPATHOGENS IN AN AGROECOSYSTEM ANT COMMUNITY 659

2006, Perfecto and Vandermeer 2008, Rietkerk and ? 2) How does habitat complexity, in this case van de Koppel 2008, Vandermeer et al. 2008). proximity of other shade trees to A. instabilis nest- Here we address the indirect effects of a fungal sites, inßuence the response of A. instabilis to a local entomopathogen, Lecanicillium lecanii (Zimmer- epizootic? and 3) How is the arboreal foraging ant man), and its ecological role, in promoting spatial community affected by the response of A. instabilis, self-organization and pattern formation. In our study the dominant ant competitor, to a local epizootic? We system, a tropical coffee agroecosystem, a keystone predicted that A. instabilis would abandon coffee arboreal-nesting ant, Azteca instabilis (F. Smith), bushes with large populations of infected scale insects forms a mutualism with the green coffee scale, Coccus and respond either by expanding their foraging activ- viridis (Green). The ant protects the scale from pred- ity to coffee bushes in the periphery of previously ators and parasites, and the scale provides a nutrient tended bushes, or by switching their foraging activity rich secretion in return (Vandermeer et al. 2002). The to shade trees, and the alternate resources therein, in ant colonies nest in shade trees that are planted in an close proximity to nests. Furthermore, we predicted approximately uniform distribution throughout the that if A. instabilis foraging activity shifted to shade farm, tending the scale insects that live on the coffee trees, it would impact the arboreal foraging ant com- plants below. The spatial distribution of the A. insta- munity by promoting expansion of competitively sub- bilis ant nests, which occur in clusters whose size dominant arboreal foraging ant activity into coffee distribution can be described with a fat tail distribu- bushes previously tended by A. instabilis, thereby pro- tion, is thought to arise through a process of self- viding access to previously unavailable resources (i.e., organization characterized by local expansion of clus- low density scale populations). In contrast, we ters counteracted by density-dependent mortality predicted that if A. instabilis were to expand foraging (Perfecto and Vandermeer 2008, Vandermeer et al. to peripheral coffee bushes, it would displace foraging 2008). A fungal entomopathogen, L. lecanii, attacks activity of other arboreal foraging ants, forcing them the local concentrations of scale insects that are as- to seek alternative resources. sociated with the clusters of ant nests, greatly reducing the scale populations and possibly causing density- Materials and Methods dependent control of the A. instabilis colonies (Jack- son et al. 2009), although other factors have been The study site is located at Finca Irlanda, a 280 ha, proposed (Vandermeer et al. 2008). organic coffee farm in the Soconusco region of We hypothesize that seasonal epizootics of L. lecanii Chiapas, Mexico (15Њ 11Ј N, 92Њ 20Ј W). The farm is a periodically disrupt ant foraging activity, causing dy- commercial polyculture (Moguel and Toledo 1999), namic shifts in the foraging ranges of A. instabilis as the with coffee bushes growing beneath shade trees that ants attempt to adapt to the loss of a primary food have been planted uniformly throughout the farm. As source. We further hypothesize that the magnitude of a result, shade trees as a whole are distributed in a these shifts depends on the availability of alternate statistically signiÞcant uniform manner, producing lit- resources located in neighboring shade trees (e.g., tle underlying habitat variability (Vandermeer et al. extraßoral nectaries or other scale insects). For the 2008). The dominant shade trees are comprised of purposes of this study, availability of alternate re- several Inga species, Alchornea latifolia (Swartz) and sources is approximated by proximity of shade trees to Trema micrantha (Blume) (Martinez and Peters A. instabilis nests, rather than directly measured be- 1996), some of which have extraßoral nectaries and cause of the difÞculties associated with sampling in- many contain various species of scale insects and sects in tropical forest canopies. Thus, habitat com- (Livingston et al. 2008). plexity, in the form of variation in the local abundance Site Selection and Data Collection. We selected of shade trees, may be modifying this interaction experimental sites based on the following criteria. (Armbrecht et al. 2004, Bos et al. 2008). While these First, each site had to have one or more A. instabilis mechanismsÑdisturbance and habitat complexityÑ nest, with each nest cluster being independent from are typically studied separately, in our study system nests outside the study plot. That is, potential sites the local interaction of a biotic disturbance driven by containing colonies with foragers traveling to and a fungal entomopathogen and habitat complexity may from nests outside of the proposed study site were be mediating the foraging activity and resource ac- rejected. This criterion was imposed to ensure that quisition of A. instabilis (the keystone ant), which colonies could not simply respond to L. lecanii could have implications for observed, landscape-level epizootics within treatment sites by increasing forag- spatial pattern (Perfecto and Vandermeer 2008, Van- ing on coffee bushes outside of the sites. Second, each dermeer et al. 2008). site had to have a large number of healthy scales so that To explore these hypotheses, we examined the im- the epizootic-induced death of these scales would pact of L. lecanii epizootics on the foraging behavior entail a signiÞcant reduction in the food resources of A. instabilis. We paid particular attention to A. available to the A. instabilis colonies. Third, we instabilis foraging response with respect to shade tree avoided sites in which the A. instabilis colonies were density (a proxy for habitat complexity) and asked primarily foraging in shade trees, either tending other three speciÞc questions: 1) How does A. instabilis scale insects or foraging at extraßoral nectaries, as we respond to L. lecanii epizootics and the concomitant wanted to focus on colonies whose primary carbohy- decrease in the food resource provided by the scale drate source was C. viridis. Using these criteria, we 660 ENVIRONMENTAL ENTOMOLOGY Vol. 42, no. 4

Fig. 1. Protocol for C. viridis and L. lecanii surveys. The coffee plant is assigned to one of the three pathways depending on how many scales are found in an initial count. If Ͼ50 scales are on the plant, the rightmost path is executed, which entails switching to a branch-by-branch estimate of the number of scales and the prevalence of L. lecanii.IfϽ20 scales are encountered, the leftmost branch is followed, and the total number of infected scales is recorded. Otherwise, an entire-plant estimate of L. lecanii prevalence is used to estimate the number of infected scales, as speciÞed by the center path. selected four treatment sites and four control sites tocol, adapted from Vandermeer and Perfecto (2006). ranging in size from Ϸ200 to 350 m2. The four treat- Each plant was rapidly surveyed to determine scale ment sites (Sites 1Ð4) and the Þrst control site (Con- insect abundance. If plants had fewer than 20 scales, trol 1) were surveyed in the summer of 2009. Because we categorized plants as “zero scales,” as virtually of logistical and time constraints, we were limited to every coffee plant in the farm has at least a few scale a single control site during the 2009 Þeld season, so, insects and A. instabilis foraging activity is primarily three additional control sites (Control 2Ð4) were restricted to plants with much larger scale popula- added in the summer of 2010; to control for the effects tions. For plants with between 20 and 50 scales, we of season, these sites were surveyed at the same time categorized them as “50 scales.” For plants with Ͼ50 of the same months as the 2009 surveys. scales, a four-category protocol was applied to each We used an aqueous suspension of L. lecanii conidia branch of the plant. Branches with 0Ð6, 7Ð30, 31Ð70, cultured from an infected C. viridis obtained from and in excess of 70 scales were placed in categories 1, within the farm to inoculate scale insects, thereby 2, 3, and 4, respectively. We then calculated an esti- creating an artiÞcial epizootic (Easwaramoorthy mated total scale count for the entire plant as (0 ϫ 1978). L. lecanii was identiÞed as the prominent fungal branches in category 1)ϩ(15 ϫ branches in category pathogen affecting C. viridis in this system based on 2)ϩ(46 ϫ branches in category 3)ϩ(150 ϫ branches both morphological identiÞcation (Zare and Gams in category 4) (Perfecto and Vandermeer 2006). A 2001) and DNA sequencing of infected scale insects ßowchart of the survey protocol is shown in Fig. 1. (D.J., unpublished data). After isolation from the scale We estimated the prevalence of L. lecanii while insect and culturing of conidia, conidia were mass performing the scale censuses (Fig. 1). For plants with produced via solid-state fermentation using cooked between 20 and 50 scales, we visually estimated the rice and sorghum as substrates. We then suspended percentage of scales infected with L. lecanii. Based on the resultant conidia in water, added Tween 80 sur- this estimate, the plant was placed in one of Þve fungal factant (0.1%) to the suspension, and applied it di- prevalence categories: 0% ϭ category 0, 1Ð10% ϭ cat- rectly to the scale insects using a handheld, manual egory 1, 11Ð20% ϭ category 2, 21Ð50% ϭ category 3, pump sprayer. Each coffee plant with scale insects and and Ͼ50% ϭ category 4. For plants assessed using the A. instabilis foragers was sprayed until the surfaces of four-category scale counting protocol (i.e., those all leaves were thoroughly wet to the point of dripping; plants with Ͼ50 scale insects), the same Þve fungal an average of 0.25 liters of suspension was applied to prevalence categories were applied to each branch each plant. Each site was sprayed twice to maximize individually. The number of infected scales was esti- L. lecanii-induced scale mortality: once on the morn- mated as 0, 0.05, 0.15, 0.35, or 0.75 times the total ing of 4 July 2009 and again the morning of 18 July 2009. number of scales for fungal categories 0, 1, 2, 3, and 4, The spore concentrations, Ϸ1.9 ϫ 105 spores/ml for respectively. the Þrst spraying, and 2.2 ϫ 106 spores/ml for the The abundance of A. instabilis and the identity of second application, were determined using a hemo- the most abundant arboreal foraging ant species were cytometer. noted for each coffee plant. Abundance was estimated We censused the scale insects at each site before by counting number and species of ant in a Þve minute (between 25 June and 3 July) and after (between 19 period on each coffee plant censusedÑbetween 14 July and 23 July) inoculation using the following pro- and 66 plants per site depending on the site. Abun- August 2013 MACDONALD ET AL.: ENTOMOPATHOGENS IN AN AGROECOSYSTEM ANT COMMUNITY 661 dance was then assessed using a four category proto- lation and summed these to determine the total num- col: Ͻ 10 foragers ϭ category 1, 10Ð25 ϭ category 2, ber of foragers in the entire site. Using these prein- 26Ð50 ϭ category 3, and Ͼ50 ϭ category 4. The ants oculation totals and the total number of foragers in were censused twice before and three times after the each site present in the Þnal ant census, we calculated experimental inoculation. Each census was conducted the percentage change in foragers. From these values at the same time of day at each site to control for we obtained the difference in the percentage change possible effects of time of day on foraging activity. at the control and treatment sites. We then combined Additionally, censuses were conducted under similar the coffee plants from both the treatment and control conditions (e.g., censuses were not conducted and sites into a single pool for resampling. compared between cool, overcast days and hot, sunny From this combined pool, coffee plants were ran- days) to the greatest extent possible. Only data from domly assigned to either the control or treatment site, the Þrst and last censuses were used in the analysis to resulting in simulated control and treatment groups ensure that our “post-epizootic” census was not inßu- with the same numbers of plants as the actual control enced by transient dynamics taking place during the and treatment sites. The difference in the percentage epizootic. change in foragers at these two simulated sites was Data Analysis. We used a bootstrapping with per- then compared with the actual, observed difference. mutation (resampling) approach to determine if the This procedure was repeated 10,000 times. The P value increase in the number of scales infected at the treat- was then calculated as the fraction of resamples re- ment sites was signiÞcantly greater than at the control sulting in a difference in percentage change that was sites; that is, to conÞrm that the L. lecanii inoculation as extreme or more extreme than the observed dif- signiÞcantly increased infection. Resampling (boot- ference. strapping) approaches are not dependent on assump- To estimate the food resources in shade trees (ex- tions about normality, which is useful in many eco- traßoral nectaries, other scale insects, and so forth) logical systems, and are becoming increasingly available in each site, we deÞned a shade tree resource accepted as the preferred method for a wide range of index. We made the simplifying assumption that the analyses given the availability of powerful computers accessibility of shade tree resources to an ant nest (Chihara and Hesterberg 2011). The method involves would fall off linearly with distance from the nest. the creation of synthetic control and treatment groups Therefore, the total shade tree resources available to by combining the data from the treatment sites and the n 1 control sites into a single pool and resampling ran- a nest are approximated by: ¥ ϭ Shade Tree Re- domly from this pool to assemble synthetic popula- iϭ1 di tions of the same sizes as the observed populations. source Index (STRI), where n is the number of shade The difference between the increase in the number of trees in the neighborhood of the nest and di is the infected scales at the synthetic control site and the distance between the nest and the shade tree. We used increase at the synthetic treatment site was then com- an analysis of covariance (ANCOVA) in the statistical pared with the observed difference between the con- computing program R (R Core Team 2010) to deter- trol and treatment sites. This procedure was repeated mine the effect of the STRI on the absolute change in 10,000 times, with the P value being the fraction of the number of A. instabilis foragers per plant. times the difference between the simulated treatment and control sites was as great, or greater than, the observed difference. This resampling approach deter- Results mines the probability that we would see, by chance alone, as large of a difference in the change at the two The treatment sites experienced nearly complete sites as was observed. infection in the Þnal survey, with prevalence increas- We used a similar resampling approach to deter- ing to an average of 92.1% from an initial baseline of mine if the changes in the abundances of A. instabilis 36.8%. In the control sites, the peak prevalence was foragers in the treatment sites were signiÞcantly dif- much lower, increasing to 45.2% from an initial value ferent from the changes observed in the control sites, of 11.8%. This increase in infected scales at the treat- using the percent change in the number of A. instabilis ment sites translated into an 87.5% reduction in the foragers on an entire-site basis. fraction of healthy scales as opposed to the 37.8% To more directly test the effect of inoculation, we background reduction observed at the control sites. also tested whether the absolute decrease in the num- The mean per-site increase in scales infected by L. ber of A. instabilis foragers was greater on sprayed lecanii from the Þrst (1 wk before treatment) to the plants than on unsprayed plants, that is, comparing Þnal (2 wk after treatment) survey was only those plants in the treatment sites that had C. 374.5 (SD ϭ 295.6) for Sites 1Ð4 and 55.9 (SD ϭ 71.0) viridis and A. instabilis in the preinoculation census to for the control sites, respectively. The increase was control site plants with C. viridis and foraging A. in- signiÞcantly greater at the treatment sites, in which a stabilis present in the preinoculation census. fungal epizootic was induced, compared with the con- The A. instabilis abundance resampling analyses trol sites (P Ͻ 0.001). Individual plants varied widely were performed as follows. First, for both the treat- in the number of scale insects infesting them (from 0 ment and control sites, we calculated the number of A. to over 25,000), which contributes to the large SDs instabilis foragers on each coffee plant before inocu- reported above. 662 ENVIRONMENTAL ENTOMOLOGY Vol. 42, no. 4

Fig. 2. Spatial distributions of arboreal foraging ants in coffee bushes at four experimental sites before (left column) and after (right column) L. lecanii inoculation. Gray circles represent numbers of A. instabilis foragers. Labeled white circles represent numbers of foragers of other ant species, as follows: (A) Brachymyrmex sp. 1, (B) Brachymyrmex sp. 2, (C) carinata, (D) Pheidole synanthropica, (E) Pheidole sp. 2, (F) Procryptocerus hylaeus, (G) Pseudomyrmex ejectus, (H) Pseudomyrmex gracilis, (I) Pseudomyrmex simplex, (J) Solenopsis geminata, (K) Technomyrmex sp., (L) Azteca sp., (M) Solenopsis sp. 2, (N) Pachycondyla sp., and (O) Camponotus sp. The sizes of the circles in each site correspond to one of four different abundance classes, from smallest to largest circles: Ͻ 10, 10Ð25, 26Ð50, and Ͼ50 foragers. Unlabeled white circles represent coffee bushes without ants. Solid black circles indicate nests of A. instabilis in shade trees. The gray circle with a cross in Site B indicates a nest of the ground-nesting ant Solenopsis geminata. Dimensions are in meters. Note that the axes are scaled differently to maximize separation between data points for visual clarity. August 2013 MACDONALD ET AL.: ENTOMOPATHOGENS IN AN AGROECOSYSTEM ANT COMMUNITY 663

The median total number of scale insects per plant Discussion (excluding plants with zero scales) at the treatment Our results show that the foraging activity of a and control sites before inoculation was 200 and 21 dominant arboreal ant, A. instabilis, is signiÞcantly scale insects, respectively (mean values: 1,085 and reduced by a fungal epizootic of L. lecanii infecting the 337). Median post inoculation was 50 scale insects in antÕs mutualist partner, the green coffee scale. This both the treatment and control sites (mean values: direct effect of the fungal pathogen on an important 1,051 and 275). Because of the presence of Þve plants food resource for the dominant ant in this system may in the treatment sites that were heavily infested with have indirect effects on the larger ant community scale insects (between 10,000 and 25,000 on a single through the observed reduction in foraging activity of plant) the median rather than the mean more accu- A. instabilis. While there exist qualitative differences rately reßects conditions in the Þeld. in the associated subdominant ant species between There was a contraction in the A. instabilis foraging each of the sites, both across space (e.g., between range at Sites 1 and 2 after the L. lecanii inoculation control Sites 2 and 3; between treatment Sites 1 and 2) (Fig. 2). The number of coffee bushes occupied by and over time (e.g., between Control 1 in 2009, and non-A. instabilis arboreal-foraging ants also increased Controls 3 and 4 in 2010; between the Þrst and last in both sites between the Þrst and last censuses: from surveys at treatment Site 2), there was a clear effect of 13 to 23 occupied bushes in Site 1 and from 6 to 15 in the experimentally induced fungal epizootic on the Site 2. The trends were less apparent in Sites 3 and 4. foraging activity of the dominant arboreal ant A. in- In Site 3, there was an increase in the number of coffee stabilis. It is possible that other unmeasured factors bushes tended by A. instabilis, from 19 to 28, while the (e.g., differences in temperature or precipitation be- number of bushes occupied by other ant species was tween years in the control plots; differences in com- relatively unchanged, decreasing from 19 to 18. How- petitive hierarchies of the associated subdominant ant ever, three bushes in this site were destroyed by a species in each plot; competitive exclusion of A. in- falling tree trunk just before the Þnal census. The stabilis from an alternate resource) could have inßu- portion of the tree trunk that fell contained part of an enced these results, but we could not identify any A. instabilis nest. The foragers found on the two newly other factors that differed systematically between the colonized bushes located at the left edge of the site control and treatment sites or between any given came from this fallen nest. In Site 4, there was a slight individual sites. What is clear is that A. instabilis for- increase in the number of coffee bushes tended by A. aging activity changed very little from the Þrst to the instabilis, from 29 to 31. However, the spatial extent of last survey in all of the control sites as compared with the A. instabilis foraging range appears to have de- the treatment sites, suggesting the induced epizootic creased slightly (Fig. 2). The number of bushes oc- is the most parsimonious explanation for the change in cupied by other ant species increased markedly, from foraging activity. 19 to 28. In the control sites (Control 1Ð4), the qual- The magnitude of this shift in foraging activity was itative picture is one of relative stasis, although there signiÞcantly modiÞed by the STRIÑor amount of al- was some change in the number of occupied plants ternative resources located in nearby shade treesÑ (Fig. 3). and was negatively correlated with foraging activity of On a per-site basis, the decrease in A. instabilis was other species of ants. These results suggest that A. greater in the treatment sites, but not signiÞcantly so instabilis colonies are able to adapt to L. lecanii (P ϭ 0.153). However, if the site where the shade tree epizootics by shifting their foraging activity, but only housing the A. instabilis nest fell (Site 3) is excluded if sufÞcient alternative resources are available. In the from the analysis, the percentage decrease in A. in- absence of alternative resources, or foraging refugia, stabilis foraging is signiÞcant (P ϭ 0.011). The de- as illustrated in the shade tree resource index, colonies crease in the abundance of A. instabilis foragers was are forced to continue tending the original, decimated signiÞcantly greater on sprayed plants than in the scale populations. Thus, colonies without foraging unsprayed plants (P ϭ 0.02). This relationship was refugia almost certainly experience a substantial re- ϭ signiÞcantly modiÞed by the STRI (ANCOVA; F3,112 duction in food intake as a result of an epizootic. In the 8.1159; P ϭ 0.005), with a larger decrease in the num- long-term, this could lead to the forced migration or ber of foragers on sprayed plants with increasing STRI mortality of these colonies, thereby contributing to (Fig. 4). the density-dependent control that gives rise to the The absolute changes in the number of A. instabilis spatial self-organization of A. instabilis nests in this foragers for Sites 1Ð4 was Ϫ360, Ϫ255, 225, and Ϫ388, system. respectively; the changes in the control sites (1Ð4) These results are counter to the hypothesis that A. were Ϫ35, 65, Ϫ55, and Ϫ18 (Fig. 5). The change in instabilis can respond to epizootics of L. lecanii by the number of coffee plants occupied by other arbo- expanding its foraging activity to coffee bushes on real-foraging ant species was negatively correlated the periphery of the original foraging range. Had A. with the change in the number of A. instabilis foragers instabilis responded by expanding foraging activity (R2 ϭ 0.86; P ϭ 0.0004; Fig. 5). That is, sites with a to the small populations of uninfected scales in pe- larger decrease in the number of A. instabilis foragers ripheral coffee bushes, the negative implications of tended to have a larger increase in the number of epizootics for the survival of the colony would likely coffee plants tended by another ant species. be mitigated. Taking advantage of these small aggre- 664 ENVIRONMENTAL ENTOMOLOGY Vol. 42, no. 4

Fig. 3. Spatial distributions of arboreal foraging ants in coffee bushes at the control sites. Gray circles represent numbers of A. instabilis foragers. Labeled white circles represent numbers of foragers of other ant species, as follows: (A) Brachymyrmex sp. 1, (B) Brachymyrmex sp. 2, (C) , (D) Pheidole synanthropica, (E) Pheidole sp. 2, (F) Procryptocerus hylaeus, (G) Pseudomyrmex ejectus, (H) Pseudomyrmex gracilis, (I) Pseudomyrmex simplex, (J) Solenopsis geminata, (K) Techno- myrmex sp., (L) Azteca sp., (M) Solenopsis sp. 2, (N) Pachycondyla sp., and (O) Camponotus sp. The sizes of the circles correspond to one of four different abundance classes, from smallest to largest circles: Ͻ 10, 10Ð25, 26Ð50, and Ͼ50 foragers. Unlabeled white circles represent coffee bushes without ants. Solid black circles indicate nests of A. instabilis in shade trees. Dimensions are in meters. Note that the axes are scaled differently to maximize separation between data points for visual clarity. August 2013 MACDONALD ET AL.: ENTOMOPATHOGENS IN AN AGROECOSYSTEM ANT COMMUNITY 665

Fig. 4. Scattergraph of ANCOVA results. The change in the number of A. instabilis foragers per coffee plant (post- inoculation minus pre-inoculation) versus the shade tree resource index was signiÞcantly different for coffee plants sprayed and unsprayed with the L. lecanii suspension, as indicated by the signiÞcantly different slopes of the regres- sion lines. The shade tree resource index is the sum of the n 1 inverse of the distances to the neighboring shade trees, ¥ . Fig. 5. Change in the number of plants occupied by other iϭ1 di arboreal-foraging ant species versus change in number of A. instabilis foragers (R2 ϭ 0.86; P ϭ 0.0004). Points are labeled with site numbers (C ϭ control). gations of unaffected scale insects would allow A. instabilis to begin to build up and tend substantial populations of C. viridis within range of the colony, epizootic at sites with low shade tree density. SufÞ- possibly leading to the formation of new satellite nests ciently high densities of shade trees surrounding A. in shade trees or large coffee bushes adjacent to the instabilis nest-sites appear to provide alternative re- expanded foraging range. sources, as well as potential nest-site locations, allow- However, in contrast with these expectations, we ing adaptation to the loss of scale insect resources in observed a contraction in foraging activity and re- coffee bushes. This Þnding highlights the possible im- source switching in sites with the highest shade tree portance of local habitat complexity in facilitating resources and increased intensity of foraging on a pattern formation. reduced number of healthy scales in sites with lower Furthermore, in the absence of readily available shade tree resources. The decline in A. instabilis for- alternative resources provided by adjacent shade trees aging activity on coffee plants in areas with abundant (Sites 3 and 4), A. instabilis increased intensity of alternate resources indicates that there is a signiÞcant scale-tending activities in coffee bushes affected by reduction in the quantity of scale resources on coffee the fungal epizootic. Based on observations made at plants impacted by L. lecanii. In the absence of sufÞ- these two particular sites, the number of foragers tend- cient alternative resources, this reduction may lead to ing the few remaining healthy scales drastically in- either a weakening of colonies and eventual nest creased after inducing the artiÞcial epizootic. This movement to escape the fungal pathogen, or nest suggests that a localized epizootic could cause persis- mortality, either of which would contribute to the tent stress because of deÞcient resource availability self-organization process and promote the emergence for colonies located in areas lacking sufÞcient shade of the spatial pattern of A. instabilis nest clusters ob- tree resources. This habitat complexity may also feed served at the landscape scale (Jackson et al. 2009). back to the self-organization process: in the absence of These results highlight how habitat complexity may alternate resources, A. instabilis colonies are more interact with resource availability to affect colony negatively impacted by L. lecanii epizootics, implying survival. A. instabilis did not respond equally at all sites that the density-dependent mortality of nests is mod- to the fungal epizootic and loss of resources. With ulated by habitat complexity. increasing shade tree density, and thus more readily The spatial pattern and distribution of A. instabilis available alternative resources, including extraßoral nests resulting from this dynamic interaction have nectaries and other species of scale insects, foraging been shown to promote biological control of coffee contracted in coffee bushes. In contrast, A. instabilis and ecosystem function in numerous other ways foraging shifted much less in response to the artiÞcial (Liere and Perfecto 2008, Perfecto and Vandermeer 666 ENVIRONMENTAL ENTOMOLOGY Vol. 42, no. 4

2008, Vandermeer et al. 2008, Jackson et al. 2009). For be the result of L. lecanii epizootics. By promoting example, A. instabilis-tended patches provide an en- self-organization of A. instabilis nests, this interaction emy free space for the development of the coccinellid between disturbance and habitat complexity may be beetle larvae of Azya orbigera (Mulsant) (Coleoptera: promoting diversity of arboreal foraging ants, which Coccinellidae), which are immune to A. instabilis at- has been shown to be crucial for maintaining biolog- tack (Liere and Perfecto 2008). In tending the green ical control in coffee through autonomous ecosystem coffee scale, A. instabilis also inadvertently protects function (Larsen and Philpott 2009, Philpott and Arm- the A. orbigera larvae from predators and parasitoids brecht 2006). (Liere and Perfecto 2008). This protection allows A. These Þndings have potentially signiÞcant implica- orbigera, an important predator of the green coffee tions for coffee management. This study has demon- scale, to develop into a mobile adult coccinellid and strated that local habitat complexity can act to dampen provide biological control of the scale in areas of the the effect of L. lecanii epizootics on A. instabilis col- farm that are unprotected by A. instabilis. There- onies, maintaining A. instabilis, and its important bio- fore, the patchy distribution of A. instabilis nest control services, in the system. Management of shade clusters provides not only an enemy free space for tree densities in coffee agroecosystems could poten- larval development, but also untended areas in tially be used to promote landscape-level pattern for- which A. orbigera acts as an effective control agent mation and local-level maintenance of diversity to for the green coffee scale (Liere and Perfecto 2008, maximize ant-derived biological control of coffee Perfecto and Vandermeer 2008, Vandermeer et al. pests. Furthermore, our treatment of a fungal ento- 2008, Jackson et al. 2009). mopathogen as a potentially important actor in driving In addition to the implications for self-organization spatial pattern formation and community interactions in this system, our results may signal that variability in is novel and could inform future study of these im- the number and accessibility of neighboring shade portant organisms. trees, that is, habitat complexity, could have ramiÞ- cations for ant diversity in the farm because of the role of the A. instabilisÐC. viridis association as a keystone Acknowledgments mutualism in this system. A. instabilis hemipteran- We thank B. Chilel for help in the identiÞcation of exper- tending activity creates patches of resources, such as imental and control sites in the Þeld. We thank G. Huerta scale insects and various other associated , from ECOSUR for considerable support in culturing L. lecanii which are aggregated around individual colonies. for experimentation and use of her laboratory facilities. We Overßow of these resources attracts numerous other thank two anonymous reviewers as well as I. Perfecto, J. species of insects including many other arboreal for- Vandermeer, S. Philpott, A. Larsen, and J. MacDonald for aging ants. The dynamic expansion and contraction of providing many useful comments on earlier drafts of the A. instabilis foraging activity mediated by L. lecanii manuscript. Funding was provided by the Rackham Grad- uate Student Research Grant, University of Michigan to epizootics and habitat complexity (availability of A.M. and D.J.; the School of Natural Resources Thesis shade tree resources) may be promoting the aggre- Research Grant, University of Michigan to A.M.; the De- gation of diversity through disruption of the compet- partment of Ecology and Evolutionary Biology Block itive exclusion that would be expected to occur in a Grant to D.J. static system. Therefore, the local contraction of the foraging range of A. instabilis, caused by the L. lecanii epizootic and mediated by the availability of foraging References Cited refugia, may promote a local increase in the foraging Alados, C. L., A. El Aich, B. Komac, Y. Pueyo, and R. activity of other arboreal-foraging ant species, includ- Garcı´a-Gonzalez. 2007. Self-organized spatial pat- ing twig-nesting and ground-nesting ants. Without this terns of vegetation in alpine grasslands. Ecol. Model. periodic disturbance, diversity of arboreal foraging 201: 233Ð242. ants would be expected to be much lower in the farm Armbrecht, I., I. Perfecto, and J. Vandermeer. 2004. Enig- as a whole. Thus, L. lecanii epizootics, in concert with matic biodiversity correlations: ant diversity responds to the variable availability of foraging refugia, may con- diverse resources. Science 304: 284Ð286. Bascompte, J., and Jordano, P. 2007. Plant- mutualis- tribute to both the generation of spatial structure tic networks: the architecture of biodiversity. Ann. Rev. through self-organization and the maintenance of ant Ecol. Evol. Syst. 38: 567Ð593. biodiversity in this system. While our results are sug- Bascompte, J., and R. Sole´. 1998. Spatiotemporal patterns in gestive of such a pattern, further research is necessary nature. Trends Ecol. Evol. 13: 173Ð174. to conclude that ant diversity is in fact increased by Bos, M. M., J. M. Tylianakis, I. Steffan-Dewenter, and T. this periodic disturbance. Tscharntke. 2008. The invasive and the In summary, our results suggest that the localized decline of forest ant diversity in Indonesian cacao agro- dynamic interaction between a biotic disturbance (L. forests. Biol. Invasions 10: 1399Ð1409. lecanii epizootics) and habitat complexity (density of Caley, M., and J. St. John. 1996. Refuge availability struc- tures assemblages of tropical reef Þshes. J. Anim. Ecol. 65: shade trees) may promote pattern formation at the 414Ð428. landscape scale. This interaction has important impli- Chesson, P. 2000. Mechanisms of maintenance of species cations for the spatial distribution of nest clusters of diversity. Ann. Rev. Ecol. Syst. 31: 343Ð366. the keystone species, A. instabilis, and provides fur- Chihara, L. and T. Hesterberg. 2011. Mathematical statistics ther evidence that the self-organization of nests may with resampling and R. Wiley, Hoboken, NJ. August 2013 MACDONALD ET AL.: ENTOMOPATHOGENS IN AN AGROECOSYSTEM ANT COMMUNITY 667

Connell, J. 1978 Diversity in tropical rain forests and coral pothenemus hampei) in southern Mexico. Agric. Ecosyst. reefs. Science 199: 1302Ð1310. Environ. 117: 218Ð221. Easwaramoorthy, S., and Jayaraj, S. 1978. Effectiveness of Philpott, S., and I. Armbrecht. 2006. Biodiversity in trop- the white halo fungus, Cephalosporium lecanii, against ical agroforests and the ecological role of ants and ant Þeld populations of coffee green bug, Coccus viridis. J. In- diversity in predatory function. Ecol. Entomol. 31: 369Ð vertebr. Pathol. 32: 88Ð96. 377. Forman, R.T.T., and M. Gordon. 1987. Landscape ecology. R Core Team. 2010. R: a language and environment for Wiley, New York, NY. statistical computing. R Foundation for Statistical Com- Hansen, R. 2000. Effects of habitat complexity and compo- puting, Vienna, Austria. (http://www.R-project.org). sition on a diverse litter microarthropod assemblage. Rietkerk, M., and J. van de Koppel. 2008. Regular pattern Ecology 81: 1120Ð1132. formation in real ecosystems. Trends Ecol. Evol. 23: 169Ð Hastings, A. 1988. Food web theory and stability. Ecology 175. 69: 1665Ð1668. Ritchie, M. 2010. Scale, heterogeneity, and the structure Jackson, D., J. Vandermeer, and I. Perfecto. 2009. Spatial and diversity of ecological communities. Princeton Uni- and temporal dynamics of a fungal pathogen promote versity Press, Princeton, NJ. pattern formation in a tropical agroecosystem. Open Rohani, P., T. J. Lewis, D. Gru¨ nbaum, and G. D. Ruxton. Ecol. J. 2: 62Ð73. 1997. Spatial self-organization in ecology: pretty patterns Larsen, A., and S. Philpott. 2009. Twig-nesting ants: the hid- or robust reality? Trends Ecol. Evol. 12: 70Ð74. den predators of the coffee berry borer in Chiapas, Mex- Scanlon, T. M., K. K. Caylor, S. A. Levin, and I. Rodriguez- ico. Biotropica 42: 342Ð347. Iturbe. 2007. Positive feedbacks promote power-law Liere, H., and I. Perfecto. 2008. Cheating on a mutualism: clustering of Kalahari vegetation. Nature 449: 209Ð212. indirect beneÞts of ant attendance to a coccidophagous Scholte, E., K. Ng’habi, J. Kihonda, W. Takken, K. Paaijmans, coccinellid. Environ. Entomol. 37: 143Ð149. S. Abdulla, G. F. Killeen, and B.G.J. Knols. 2005. An Livingston, G. F., A. M. White, and C. J. Kratz. 2008. Indi- rect interactions between ant-tended hemipterans, a entomopathogenic fungus for control of adult African dominant ant Azteca instabilis (: Formi- malaria mosquitoes. Science 308: 1641Ð1642. cidae), and shade trees in a tropical agroecosystem. En- Shah, P., and J. Pell. 2003. Entomopathogenic fungi as bio- viron. Entomol. 37: 734Ð740. logical control agents. Appl. Microbiol. Biotechnol. 61: Martinez, E., and W. Peters. 1996. La cafeticultura bi- 413Ð423. olo´ gica: la Þnca Irlanda como estudio de caso de un Skarpe, C. 1991. Spatial patterns and dynamics of woody desen˜ o agricoecolo´ gico, pp. 159Ð183. In J. T. Arriaga, et vegetation in an arid savanna. J. Veg. Sci. 2: 565Ð572. al. (eds.), Ecologõ´a Aplicada a la Agricultura: Temas Sole´, R. 2007. Scaling laws in the drier. Nature 449: 151Ð Selectos de Me´ xico. Universidad Autonomo Metropoli- 153. tana. Sole´, R. V., and J. Bascompte. 2006. Self-organization in Meyling, N. V., and A. E. Hajek. 2010. Principles from com- complex ecosystems. Princeton University Press, Prince- munity and metapopulation ecology: application to fun- ton, NJ. gal entomopathogens. BioControl 55: 39Ð54. Turner, S. J., R. V. O’Neill, W. Conley, M. R. Conley, and Moguel, P., and V. Toledo. 1999. Review: biodiversity con- H. C. Humphries. 1990. Pattern and scale: statistics for servation in traditional coffee systems of Mexico. Con- landscape ecology. In M. G. Turner and R. H. Gardner servation Biol. 13: 11Ð21. (eds.), Quantitative Methods in Landscape Ecology. Paine, R. 1966. Food web complexity and species diversity. Springer, New York, NY. Am. Nat. 100: 65Ð75. Vandermeer, J., and I. Perfecto. 2006. A keystone mutual- Paine, R. 1969. The Pisaster-Tegula interaction: prey ism drives pattern in a power function. Science 311: 1000Ð patches, predator food preference, and intertidal com- 1002. munity structure. Ecol. 50: 950Ð961. Vandermeer, J., I. Perfecto, G. I. Nun˜ ez, S. M. Philpott, and Pascual, M., M. Roy, F. Guichard, and G. Flierl. 2002. Clus- A. G. Ballinas. 2002. Ants (Azteca sp.) as potential bio- ter size distributions: signatures of self-organization in logical control agents in shade coffee production in spatial ecologies. Philos. Trans. R. Soc. B: Biol. Sci. 357: Chiapas, Mexico. Agroforestry Syst. 56: 271Ð276. 657Ð666. Vandermeer, J., I. Perfecto, and S. M. Philpott. 2008. Clus- Pell, J. K., J. J. Hannam, and D. C. Steinkraus. 2010 Con- ters of ant colonies and robust criticality in a tropical servation biological control using fungal entomopatho- agroecosystem. Nature 451: 457Ð459. gens. BioControl 55: 187Ð198. Zare, R., and W. Gams. 2001. A revision of Verticillium sec- Perfecto, I., and J. Vandermeer. 2008. Spatial pattern and tion Prostrata. IV. The genera Lecanicillium and Simpli- ecological process in the coffee agroforestry system. cillium gen. nov. Nova Hedwigia 73: 1Ð50. Ecology 89: 915Ð920. Perfecto, I., and J. Vandermeer. 2006. The effect of an ant- hemipteran mutualism on the coffee berry borer (Hy- Received 8 October 2012; accepted 10 June 2013.